This thesis comprehends four essays investigating strategies to fight against poverty. The first essay explores a series of police operations to pacify the slums of Rio de Janeiro to understand the impacts of intrauterine exposure to violence on birth outcomes. One argues that pregnancies starting before, but ending around the pacification dates are ‘quasirandomly’ exposed to exogenous shocks of violence during pregnancy. The results show that each month pregnant women are exposed to pacification increases birth weights by 4 grams and reduces the probability of low birth weight (< 2500 grams) by 1.2 percent compared to pregnancies ending just before pacifications. A second essay uses Brazilian legislative change making it mandatory for private hospitals to publicly disclose information about physicians’ performance. The results show a reduction in scheduled C-sections by 4.8 percent; which two-thirds originating from physicians anticipating to information disclosure. The third essay proposes an empirical strategy to estimate bullying effects on labour and schooling outcomes when "true" bullying is observed inaccurately. The estimates show that high-school bullying decreases University attendance by 3.4 percent and increases the probability of being not in education, employed or in training after high-school by 2.8 percent. Estimations neglecting misreport implicates in impacts two-thirds smaller. And finally, the fourth essay shows that poor households increase their participation in social groups after receiving Bolsa Família. The strategy explores households registered in Cadastro Único, and performs a propensity score difference-in-difference framework to minimize selection bias. Becoming a recipient of Bolsa Família increases .09 standard deviations the number of social affiliation and increase from 6.1 to 8.9 percent the probability of engaging in social groups. Altogether, this thesis implicates that investing in early stages of life harvest significant benefits to disadvantaged children, it also shows that victims of bullying need sustained support after high school, and that conditional cash transfers foster social engagement.

This dissertation consists of three studies in the field of health economics. The first chapter studies the market situation of the U.S. nursing home industry. It uses the most recent data available from the Annual Survey of Nursing Homes conducted in Wisconsin. In this study, we derive theoretical predictions from an optimization problem of a representative nursing home under various assumptions. We introduce a new measure, a home's bed-utilization rate, in our empirical strategy and find evidence of excess demand from Medicaid patients in Wisconsin. A positive relationship between Medicaid payment rates and private-pay prices is found in homes with high bed utilization. Additionally, we find strong adverse effects of higher reimbursement rates on quality measures. These findings prove there is an excess demand from Medicaid patients in Wisconsin. This conclusion has direct implications for the quality of care that a nursing home provides for its patients. The second study takes advantage of the "natural experiment" features of the major health care reform in Thailand in 2002 in order to estimate the price elasticity of health care demand among Thai citizens. We use the difference-in-difference technique to capture the pure effect of the reform on the health care utilization behavior of those who were directly affected by the reform. In order to capture any secular trend in health care utilization, we use data from a group of people who were not affected by the reform. We find that the reduction in health care price immediately induced those who lacked health insurance coverage to increase their visits to a public health care facility, although similar trends were not found a few years after the reform. The estimated change in visits is used to calculate the price elasticity of demand, which falls in the range of -1.36 to -0.58. The last study examines the relationship between risky behaviors among Americans aged 50-65 and their health insurance coverage. Despite the fact that moral hazard behaviors are predicted by economic theory, the study finds that health insurance has no significant effect on certain risky behaviors such as smoking. Surprisingly, we find a significantly positive relationship between health insurance coverage and healthy behaviors such as exercising regularly. This finding reflects the importance of health insurance companies in providing its customers with more health information that could encourage health-oriented attitudes. / text

No description available.

This thesis contributes new knowledge to discussions of inequality in three arenas and two methodological syntheses that might inform future statistical analyses. Methodologically, the application of unconditional quantile regression in a two-stage model is used to determine whether response bias plays any role in the patterns observed in survey responses (Chapter 2), and, a recent development in the program evaluation literature (the synthetic control method) is combined with flexible parametric survival models to identify treatment effects where stratification is perfectly correlated with treatment (albeit under restrictive assumptions). The analyses undertaken herein have discovered: that self-assessed unmet need for healthcare has an empirical basis for application as reporting behaviour statistically predicts decline in health, that the likelihood of reporting unmet need conditional on health and healthcare utilization is correlated with the dimensions along which social scientists might map inequality, that government programs intended to provide a minimum level of utility are unresponsive to regional poverty-relief efforts, and that household bargaining outcomes regarding number of children can be predicted by exposure to a parental divorce. The implications of these findings are manifold. First, while self-assessment of healthcare access is a valid metric on average to overcome limitations of needs-adjusted utilization, its use in cross-sectional analysis as it is currently obtained in survey across many different jurisdictions is suspect. Second, the patterns of fertility conditional on parents’ divorce suggest that household bargaining in Canada does not likely belong to several different theoretical frameworks. Specifically, bargaining most likely exists in an environment where women still bear the cost of children in the event of a divorce, or bargaining exists without commitment. Finally, while the theoretical literature makes compelling claims about interactions between different levels of government policy, in practice this may not be the case even if policy-wordings seem to suggest this would be particularly relevant. / Thesis / Doctor of Philosophy (PhD) / This thesis explores three separate dimensions of inequality. First, a method of improving measurement of inequity in healthcare is demonstrated in a world of heterogeneous preferences where traditional methods exploiting observed utilization are shown to be inadequate. Potential issues resulting from response bias in the metric used in the method from chapter 1 are investigated in chapter 2. Next, the experience of a parents’ divorce as a child is correlated with adult fertility showing that the intergenerational transmission of marital instability may influence decisions on family size as an adult, specifically, only women show a change in fertility outcomes after the exposure to their parents’ divorce. Finally, the effect of a regional transfer intended to improve living standards for the poor is examined for its effect on the workfare program in Ontario. The transfer is found to increase the duration of welfare benefit receipt by two months, representing a welfare improvement for eligible recipients.

This dissertation consists of three research papers in Applied Microeconomics. Each paper uses an econometric technique to analyze a problem related to human behavior. The first paper examines the separate effects of time and location of the School Breakfast Program on participation and consumption of breakfast by elementary school children in northern Nevada. Controlling for potential selection bias and unobserved individual fixed effects with a panel version of the Heckman sample selection model, it is shown that extra time allowed for breakfast leads to an approximately 20% increase in average participation, and the transition from cafeteria to classroom adds another 40% for the typical student. The second paper uses the Hedonic Property Valuation Method to quantify the willingness-to-pay of residents in the Dan River region for three dimensions of an improved food environment---availability, accessibility, and acceptability of food. This paper accounts for potential omitted variables issue in the hedonic analysis by applying a spatial-lag model, and finds an overall negative or null preference of residents in this region for an improved food environment. The third paper investigates the effects of characteristics of human interpreters and images on the accuracy of cloud interpretation for satellite images in an online experiment, using a fractional logit model. The results indicate that an image with higher cloud coverage and/or larger brightness is more likely to receive higher accuracy, and the more time spent on the image and more image completed are also beneficial for improving the accuracy. This paper also uses a logistic regression model to compare the performance of human interpreters to that of an automated algorithm, and finds that human interpreters outperform the automated algorithm for an average satellite image out of our twelve selected images. / Ph. D.

Esta dissertação estuda padrões sociais e raciais de nomes e sobrenomes entre alunos paulistas e a correlação desses padrões com desempenho escolar. Para cada nome e sobrenome observados criamos índices que medem o quão distintamente rico (ou pobre) e o quão distintamente branco (ou afro) tais nomes e sobrenomes são. Os resultados são de que tanto nomes quanto sobrenomes dão sinais de status socioeconômico, enquanto sobrenomes e, em menor medida, nomes predizem raça. Nomes que indicam maior status são, em geral, de maior frequência estatística, de origem latina, formados por um só termo e com grafia condizente com a Língua Portuguesa formal. Nomes de menor frequência estatística, compostos por duas palavras, com influência do idioma inglês na grafia e na pronúncia e com grafias distintas do Português formal estão associados a baixo status. A evidência aponta ainda para uma relação mais forte entre primeiro nome e status do que entre primeiro nome e raça, revelando que, em São Paulo, o universo cultural que baseia a escolha de primeiros nomes é mais classe-específico que raça-específico. Pelo lado dos nomes de família, os dados mostram que sobrenomes tipicamente de alto status e tipicamente brancos são de menor frequência estatística e de origem não-portuguesa. Sobrenomes portugueses não apresentam padrão socioeconômico e racial, com exceção dos três sobrenomes mais frequentes, os quais são tipicamente pobres. Além disso, sobrenomes associados à religião católica são tipicamente pobres e afrodescendentes. Posteriormente, avaliamos a relação entre diversas medidas de performance escolar e nomes e sobrenomes. Os dados apontam para uma associação robusta entre ter um nome distintamente pobre (rico) e um sobrenome distintamente afro (branco) e piores (melhores) resultados escolares. Tal evidência é consistente com um cenário em que escolhas culturais dos pais e herança cultural familiar afetam a acumulação de capital humano de crianças e jovens adultos. Entre outras possibilidades está a existência de tratamento discriminatório nas escolas. / This dissertation studies social and racial patterns of names and surnames for students in São Paulo and evaluates its contribution to academic performance. For each name and surname observed, we create indices to measure how distinctively rich (or poor) and how distinctively white (or black) they are. We find that both names and surnames are predictors of social status, while surnames and, to a lesser extent, names predict race. Names that indicate higher status present, generally, higher frequency, latin origin, only one term and spelling coherent to formal Portuguese language. Names that signal low status show lower frequency, two terms, English language influence on spelling and pronouncing and spelling different from formal Portuguese. The evidence also points to a stronger link between first names and social status than between first names and race, revealing that cultural determinants of name choice are mainly social rather than racial. When it comes to family names, data shows that distinctively high status and white surnames have low frequency and non-portuguese origin. Portuguese surnames do not present social or racial patterns, except for the three most common surnames, which are typically poor. Additionaly, catholic devotional surnames are distinctively poor and afro. Then we evaluate the correlation between names, surnames and a set of academic performance measures. The evidence indicates a robust link between having a distinctively poor (rich) name and a distinctively afro (white) family name and worse (better) grades. This is consistent with a scenario where parental cultural choices and familiar cultural heritage affect human capital accumulation of children. Discriminatory treatment in school is another possibility.

<div>The following dissertation presents three, stand-alone chapters on incentives and organization in the analysis of public policy. The first two chapters use administrative data from the court system of North Carolina to (1) provide evidence of strategic scheduling decisions for felony cases around judicial rotations and (2) evaluate a wide range of alternative measures of judicial severity, a common methodology used in random judge assignment for evaluation of sentencing on recidivism and in the broader field of scoring and third-party evaluation. The first chapter presents a variety of tests for strategic scheduling, finding systematic variation in the disposition of cases by judges in their election district and by defendant gender. The second chapter presents coefficients from two-stage least squares estimates of the effect of incarceration, probation, and sentence length on recidivism, finding broadly consistent results in direction, but with a significant degree of variation in point estimates and confidence interval size. In addition, alternative indexes sometimes pass and sometimes fail balance and monotonicity tests. Finally, evidence is presented that there are multidimensional elements to judicial propensities around sentence type and sentence intensity, indicating that a single severity index may miss important variations in judge types that are meaningful to defendant outcomes. The third chapter uses data from the 1992 to 2017 Censuses of Agriculture to evaluate the impact of the Land Contract Guarantee Pilot Program (LCGP) on the share of new farmers in counties across the United States. Estimates of shares of new farmers from a difference-in-differences model are then used to assess the impact of new farmers on aggregate measures of farm capital and federal program participation. In general, the LCGP has a significant and positive effect on the share of new farmers. Counties with higher shares of new farmers are less likely to participate in federal conservation programs and have lower total machinery assets. An event study approach that includes data from 2017, when the LCGP was expanded nationwide, confirms positive effects from the program, but offers conflicting views on farm capital and program participation. It is suggested that this is likely a result of additional programs put in place in the 2014 Farm Bill, and future research is proposed to address this conflicting sign.</div>

Esta dissertação estuda padrões sociais e raciais de nomes e sobrenomes entre alunos paulistas e a correlação desses padrões com desempenho escolar. Para cada nome e sobrenome observados criamos índices que medem o quão distintamente rico (ou pobre) e o quão distintamente branco (ou afro) tais nomes e sobrenomes são. Os resultados são de que tanto nomes quanto sobrenomes dão sinais de status socioeconômico, enquanto sobrenomes e, em menor medida, nomes predizem raça. Nomes que indicam maior status são, em geral, de maior frequência estatística, de origem latina, formados por um só termo e com grafia condizente com a Língua Portuguesa formal. Nomes de menor frequência estatística, compostos por duas palavras, com influência do idioma inglês na grafia e na pronúncia e com grafias distintas do Português formal estão associados a baixo status. A evidência aponta ainda para uma relação mais forte entre primeiro nome e status do que entre primeiro nome e raça, revelando que, em São Paulo, o universo cultural que baseia a escolha de primeiros nomes é mais classe-específico que raça-específico. Pelo lado dos nomes de família, os dados mostram que sobrenomes tipicamente de alto status e tipicamente brancos são de menor frequência estatística e de origem não-portuguesa. Sobrenomes portugueses não apresentam padrão socioeconômico e racial, com exceção dos três sobrenomes mais frequentes, os quais são tipicamente pobres. Além disso, sobrenomes associados à religião católica são tipicamente pobres e afrodescendentes. Posteriormente, avaliamos a relação entre diversas medidas de performance escolar e nomes e sobrenomes. Os dados apontam para uma associação robusta entre ter um nome distintamente pobre (rico) e um sobrenome distintamente afro (branco) e piores (melhores) resultados escolares. Tal evidência é consistente com um cenário em que escolhas culturais dos pais e herança cultural familiar afetam a acumulação de capital humano de crianças e jovens adultos. Entre outras possibilidades está a existência de tratamento discriminatório nas escolas. / This dissertation studies social and racial patterns of names and surnames for students in São Paulo and evaluates its contribution to academic performance. For each name and surname observed, we create indices to measure how distinctively rich (or poor) and how distinctively white (or black) they are. We find that both names and surnames are predictors of social status, while surnames and, to a lesser extent, names predict race. Names that indicate higher status present, generally, higher frequency, latin origin, only one term and spelling coherent to formal Portuguese language. Names that signal low status show lower frequency, two terms, English language influence on spelling and pronouncing and spelling different from formal Portuguese. The evidence also points to a stronger link between first names and social status than between first names and race, revealing that cultural determinants of name choice are mainly social rather than racial. When it comes to family names, data shows that distinctively high status and white surnames have low frequency and non-portuguese origin. Portuguese surnames do not present social or racial patterns, except for the three most common surnames, which are typically poor. Additionaly, catholic devotional surnames are distinctively poor and afro. Then we evaluate the correlation between names, surnames and a set of academic performance measures. The evidence indicates a robust link between having a distinctively poor (rich) name and a distinctively afro (white) family name and worse (better) grades. This is consistent with a scenario where parental cultural choices and familiar cultural heritage affect human capital accumulation of children. Discriminatory treatment in school is another possibility.

The central goal of my dissertation is to answer important questions about market design in health care when consumers have inertia, using modern industrial organization tools. The presence of consumer inertia in several markets has been well established in the literature, although we still know very little about how inertia affects the way markets work. In my dissertation, I shed light on these issues in the context of different institutional settings of health care sectors in different countries. Health care markets are extremely relevant because of their huge impacts on the quality of life and on mortality of individuals. In times when the expenditure on health care is increasingly high in modern economies, a better understanding of how these markets work is needed in order to decrease costs and improve their performance. The first chapter disentangles the effects of reductions in switching costs and in the length of contracts (lock-in) on consumer welfare, using quasi-experimental variation in the length of contracts in the Uruguayan health care system. In the second chapter, I study the effect of supply-side firm responses in terms of pricing and offering of new products, on consumer welfare in Medicare Part D in the U.S. Finally, the third chapter studies the effects of increased competition induced by reductions of consumer inertia, on quality and returns to skills for physicians, using uniquely detailed data from the Uruguayan health care sector. The use of tools from the field of industrial organization allows me to combine a solid theoretical background with clearly identified reduced-form and structural models, to analyze the welfare implications of equilibrium behavior in these markets, and to evaluate policy interventions and regulations aimed at improving welfare.

This dissertation consists of three separate research papers. Each paper uses a different econometric technique to analyze a problem relating to the social aspects of climate change. The first paper investigates a potential adaptive strategy to counteract warming stream waters through stream intervention projects. Using novel non-parametric matching estimation techniques it is shown that these intervention projects have positive effects on homeowners that are near to the stream but downstream of the project site. The second paper uses Bayesian econometric techniques to analyze survey data regarding the welfare losses experienced as a result of power outages across Europe. This paper shows how the severity and spatial distribution of these welfare losses will change as the climate warms, which enables the current electricity grid expansion taking place in Europe to account for these effects of climate change. The third paper uses Classical econometric techniques to estimate the effect of temperature on visitor recreation choices around Lake Tahoe. It is then shown that under climate scenarios the demand for beach and water access at Lake Tahoe will greatly increase, which suggests that lake managers begin to plan regulations and build infrastructure to account for this demand increase. / Ph. D.